At 07:07 AM 9/17/2019, Brian Sheppard via Computer-go wrote:
>I remember a scheme (from Dave Dyer, IIRC) that indexed positions based on the
>points on which the 20th, 40th, 60th,... moves were made. IIRC it was nearly a
>unique key for pro positions.
Correct, but it's only useful
At 06:31 AM 3/5/2018, valky...@phmp.se wrote:
>My guess is that there is some kind of threshold depending on the relative
>strength of MC eval and the value function of the NN.
My experiments suggest it's better to train with much longer MCTS
time than will be used in actual play, so the MCTS
The exact meaning of the result MCTS returns is irrelevant. The
net should just learn it.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
At 08:05 AM 12/8/2017, Jim O'Flaherty wrote:
>Dave,
>
>To whom is the "your" in your first sentence referring? There is no context
>from which to derive to whom you are speaking.
Sorry, it was "your" reference to Abchij, but I intended it to be
about any attempt to make games that are difficult
Without reference to your specific ideas for games that might be
difficult to solve, I wonder where these games fit on the human
playability scale. The things we find acceptable as games are
in a pretty small domain, which lies between the things that are
trivial and the things that are too
My question is this; people have been messing around with neural nets
and machine learning for 40 years; what was the breakthrough that made
alphago succeed so spectacularly.
___
Computer-go mailing list
Computer-go@computer-go.org
While MCTS works "better" in games with a forward direction, it eventually
converges to the same answer as alpha-beta in any game. The general
architecture is to set a maximum depth, and use a suitable evaluator
at the leaf nodes.
I haven't done detailed studies, but there is definitely a
While MCTS works "better" in games with a forward direction, it eventually
converges to the same answer as alpha-beta in any game. The general
architecture is to set a maximum depth, and use a suitable evaluator
at the leaf nodes.
I haven't done detailed studies, but there is definitely a
I wonder how alphago-0 treats the menagerie of special positions, such as
bent 4 in the corner, thousand year ko, rotating ko, etc.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
I wonder how alphago-0 treats the menagerie of special positions, such as
bent 4 in the corner, thousand year ko, rotating ko, etc.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
The tree built by MCTS is very unbalanced - some branches
are explored more thoroughly than others.
Tweaking the algorithm to favor the newer results might result
in an overall improvement, but it also would be subject to all
the pitfalls of partially or unevenly evaluated trees.
This is
The tree built by MCTS is very unbalanced - some branches
are explored more thoroughly than others.
Tweaking the algorithm to favor the newer results might result
in an overall improvement, but it also would be subject to all
the pitfalls of partially or unevenly evaluated trees.
This is
Character encoding (usually UTF8 these days) ought not to be part of
the standard, it ought to be up to the containing file to describe the
encoding at that level.Likewise, nothing in the standard ought to
require support for particular character sets. Rather, if a sgf record
contains an
Baduk cap does a reasonably good job of recognising where the stones
are, but that's a far cry from recording a game as it progresses. It
doesn't do life and death of the endgame position.
___
Computer-go mailing list
Computer-go@computer-go.org
At 11:31 AM 6/10/2016, uurtamo wrote:
>Compiler no workie? ;)
Lazy. Lots of junk to install and get working.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
Now if someone would post a binary that would just run on suitable hardware.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
Now if someone would post a binary that would just run on suitable hardware.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
A lot of what passes through this list would be just noise on a forum, but
it's good to have a public face for summaries, announcements, and items
with crossover interest to other games.
___
Computer-go mailing list
Computer-go@computer-go.org
I was recently working on assigning final scores to completed games, using
the large data set from Badukmovies.com.
My observation is that the size of the data set (50,000 games) is not
large enough to get good coverage of unusual situations occurring in real
games.
There's a definite need
I was recently working on assigning final scores to completed games, using
the large data set from Badukmovies.com.
My observation is that the size of the data set (50,000 games) is not
large enough to get good coverage of unusual situations occurring in real
games.
There's a definite need
>Can you tell us the rules of the game? Maybe they help to explain the
>phenomenon.
The game is https://boardgamegeek.com/boardgame/149910/six-making
The most unusual thing I see in the UCT tree is that at all the moves
seem to be evaluated about the same, right up to the point where wins
Developing a UCT robot for a new game, I have encountered a
surprising and alarming behavior: the longer think time the
robot is given, the worse the results. That is, the same robot
given 5 seconds per move defeats one give 30 seconds, or 180 seconds.
I'm still investigating, but the
At 10:59 AM 11/6/2015, Gonçalo Mendes Ferreira wrote:
>That doesn't seem very realistic.
This is with a well tested framework that's been used for 20+ games.
Whatever the ultimate resolution, the counter intuitive result that
triggered it stands alone; longer think times give worse results.
I
>But you are stopping right at the atari, and then pile on playouts that make
>it seem work...
Yes, something like that may be the situation that turns the result.
Suppose the tree stops at a point where there are two moves, a
blunder that leads to a quick end, and another move which leads
to
>I have seen this exact behavior when first experimenting with long thinking
>times in Pachi. When you stop growing the tree, the algorithm degenerates to
>"delayed" single-level Monte Carlo along the principal variations, with all
>the MC-without-tree weaknesses.
The pathology definitely
How about handicapping the hardware based on time. Programs running
on more powerful hardware would get less time.
On the other hand, improving the software includes making use of more
powerful hardware. Handicapping (or banning) powerful hardware would
discourage that.
On third hand,
Maybe, maybe not. Current research in volition and conscious choice
indicates that conscious choice is actually an after the fact explanation
of decisions based on unconscious processes.
I found this video about how research into visualizing neural networks
led to big improvements, by showing
My reading of the material currently available is that the big advance was
to forward prune a lot of moves based on simple utility metrics.
Moves that don't do something are thrown out.
A reasonable idea, but the utility metrics are specific to arimaa;
and even for the general idea, it's hard
It's easy to get 20+ ppl saying OMG I want to help. But I need at
least 1-2 more core devs outside me, especially if I'm working in a
language outside my zone.
All I can say is, if you need 1-2 outside collaborators, you better
have a plan B. Everyone dances to their own tune, and no one
Is anyone else bothered the line endings of cgos messages are being
transformed, which makes everything mash onto one long line on
my mail reader ?
___
Computer-go mailing list
Computer-go@computer-go.org
I'm planning to add Go (for human players) to Boardspace.net soon, and
I would like to have at least a minimal house robot to play on small boards.
It's absolutely required that the bot be written in Java. Other than that,
the main requirement is cooperation to integrate the bot into the game
Anyway, it's very easy to make a fast PRNG these days.
A couple words of caution about hacking PNRG.
Back in the stone age of computing, some mavens from the triple-i movie
group cooked up a galaxy simulator which generated pictures of spiral
galaxies based on a numerical model. The
Read and Write is easy, I have some classes I use for all my games
that I've distributed several times.
Edit implies a full gui, which is a little more complicated. I have
a fairly primitive board widget I've distributed in the past.
If your goal is a fully featured go gui, with replay and
At 01:14 PM 3/1/2015, folkert wrote:
Hi, It looks like cgos.computergo.org is down?
the server has moved back to cgos.boardspace.net
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
You don't need a neural net to predict pro moves at this level.
My measurement metric was slightly different, I counted how far down the
list of moves the pro move appeared, so matching the pro move scored
as 100% and being tenth on a list of 100 moves scored 90%.
Combining simple metrics such
(i) IGS is derivation of NNGS, which is free software (GPLv2)! It has
even seen some slight development in past few years.
I don't think that's correct - NNGS was a functional copy of IGS created
by duplicating the published (telnet based) interfaces. It eventually
was open sourced before it
At 10:12 AM 10/24/2009, Joshua Shriver wrote:
Came across this today, and since this is also an AI oriented list thought
some of you might enjoy it too.
In any rating scheme, who you play can be as important as how well.
This is especially true for small groups.
Suddenly adding or dropping a strong player will certainly cause
all the other player's ratings to shift.
___
computer-go mailing list
It's easy to construct self-atari of unlimited size that both
can occur and should be played, if the capturing move that follows
the self-atari is then recaptured in a snapback.
___
computer-go mailing list
computer-go@computer-go.org
If you are in a lost position, good play is play that maximizes
the probability of a turnaround, which is quite different depending
on how far behind you are, and for what reason.
If the status of all the major groups is solid, then concentrating
on tactics which can gain a few points reliably
My $0.02
The choice of language is mostly arbitrary.
CGOS is really two separate programs:
(1) an I/O multiplexer that manages the clients connections and detailed
communication,
(2) a scheduler/planner/recorder that manages the overall operation of the
site.
I would definitely separate
So I believe this is a design flaw in CGOS itself. I wrote CGOS without
having had any experience writing servers.
If there's a problem with larger databases, perhaps it can be
fixed by adding the right indexes to the sql database. If
you add a little time monitoring code around your
And somehow I don't ever see comments anywhere suggesting that this could be a
problem. So what I'd like to know is: is this so trivial that no one ever
mentions it, or are the heuristics that programs use to terminate playouts so
obscure that they are too embarrasing to mention?
Completely
I've written dozens of games with alpha-beta searches, so I think
it's fair to say that I have a basic understanding of the process.
Your description is correct but incomplete. Alpha beta is good at eliminating
lines of play once a strong outcome is known somewhere in the tree, but much
weaker
Some lines of play involving large captures will effectively never
terminate, even with superko rules in effect.
I doubt it is possible to eliminate all these non-terminating
lines of play in any way that is provably correct.
.. So while claims of solution by exhaustive search might be very
You can just prove that you can make a large-enough chain that is
unconditionally alive. I believe that's what Erik did. In practice,
you cannot do an exhaustive search using superko rules because then
hash table scores cannot be used.
I don't think you can always do that. For example, if
At 06:31 PM 5/22/2009, David Doshay wrote:
there are no chains of size 30 on a 5x5 board,
I'll concede for a 5x5 board, but I think my point
is valid for sufficiently large boards, probably 7x7.
Almost any strategy other than playing out all legal moves
involves a lot of hand waving that is
Storing an opening book for the first 10 moves requires
331477745148242200 nodes. Even with some reduction for symmetry,
I don't see that much memory becoming available anytime soon, and you still
have to evaluate them somehow.
Actually storing a tree, except for extremely limited
At 02:13 PM 5/12/2009, Michael Williams wrote:
Where does your 99% figure come from?
1/361 1%
by endgame there are still easily 100 empty spaces
on the board.
___
computer-go mailing list
computer-go@computer-go.org
At 02:13 PM 5/12/2009, Michael Williams wrote:
Where does your 99% figure come from?
1/361 1%
by endgame there are still easily 100 empty spaces
on the board.
___
computer-go mailing list
computer-go@computer-go.org
An essential feature of monte carlo is that it's search space is
random and extremely sparse, so consequently opportunity to re-use
nodes is also extremely sparse.
On the other hand, if the search close to the root is not sparse, my
previous arguments about the number of nodes and the number of
I assume Dave Dyer does not understand alpha beta pruning either, or he would
not assume the branching factor is 361.
The branch at the root is about (361-move number) - you have to consider
all top level moves. A/B only kicks in by lowering the average branching
factor at lower levels
If I use persistent storage and do that search again in another game, I can
start exactly where I left off and generate 50,000 more nodes. It will be
the same as if I did 100,000 nodes instead of 50,000 nodes.Or put another
way, it will be the same as if I spent 20 seconds on this
But then MCTS is invalid. The point is that you do spend time learning that
these nodes are not relevant, so you might as well try to remember that.
It is invalid. It's just a heuristic that is working within the current domain.
If you are playing a game of chess and fall for a trap, do
Highly recommended
http://www.youtube.com/watch?v=EjrvrH1bKIwfeature=PlayListp=2C02F6B33145E762index=0playnext=1Mathematics
and Go by Elwyn Berlekamp
___
computer-go mailing list
computer-go@computer-go.org
Donn, your email at d...@mit.edu is bouncing.
___
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
While your goal is laudable, I'm afraid there is no such thing
as a simple tree search with a plug-in evaluator for Go. The
problem is that the move generator has to be very disciplined,
and the evaluator typically requires elaborate and expensive to
maintain data structures. It all tends to be
Do you mean that the evaluator might be used during move ordering somehow
and that generating the nodes to expand is tightly coupled with the static
evaluator?
That's the general idea.
No search program can afford to use a fan-out factor of 361. The information
about what to cut has to come
Do you mean that the evaluator might be used during move ordering somehow
and that generating the nodes to expand is tightly coupled with the static
evaluator?
That's the general idea.
No search program can afford to use a fan-out factor of 361. The information
about what to cut has to come
This is old and incomplete, but still is a starting point you might
find useful http://www.andromeda.com/people/ddyer/go/global-eval.html
General observations (from a weak player's point of view):
Go is played on a knife edge between life and death. The only evaluator
that matters is is
At 12:59 AM 2/4/2009, David Fotland wrote:
What do you mean by operator at remote end? In my case, the program was
running on a cluster at Microsoft in some computer data center. There was
no operator at Microsoft. The cluster was operated from Beijing through a
remote desktop. The operator
My theory is that the organizers of tournaments with remote participants
could appoint official observers, to observe the operators at the remote
end of connections. Not foolproof, but simple and doesn't interfere with
the conduct of the tournament.
There's already a havannah section on this game programming
forum: http://www.grappa.univ-lille3.fr/
-- which could use an influx of traffic.
___
computer-go mailing list
computer-go@computer-go.org
Lets look at it another way - no one would care what hardware
you choose to use, unless you win. So at the very least, you
ought to be able to use arbitrary hardware until it becomes
established that only that class of hardware can win.
___
I think general hardware limits are good, because they will permit
more teams to be competitive without altering the nature of the
competition.
___
computer-go mailing list
computer-go@computer-go.org
At 01:52 AM 11/27/2008, Denis fidaali wrote:
...
But what really lacks (or i wasn't able to find anyway) is a strong community
like there is for go.
A CGOS equivalent.
A GTP equivalent.
A Gogui equivalent.
A Kgs equivalent.
I don't think there's a match between your goals and what
Permit me to play the skeptic here; I think you're going about it absolutely
backwards - unless you already have a strong algorithm which depends on 128 bit
rotations, and only lack an efficient hardware engine to run it on.
If your idea of fun is to really feel the bits squishing between your
That's impressive, especially considering the fairly long search path between
Go and igowin.
It happens. One day recently I was idling at boardspace.net, when
in the course a few minutes the site was overrun by about 30 guests,
all speaking German and wanting to play Hex. It turned out that
At 01:31 PM 11/26/2008, Denis fidaali wrote:
Speaking of hex ... I really think it would be a nice intermediary game before
tackling the complexity of go. Do you know of any good community (and protocol
equivalent to GTP) where i could start to look for submitting a bot ?
There are a couple of
Here's a chance to share an amusing and illustrative anecdote.
I was working on optimizing Goodbot, a program that plays Tantrix,
and because of the nature of the game, the only way to really qualify
an improvement is to run many test games against a standard opponent.
At one point, I was
Here's a chance to share an amusing and illustrative anecdote.
I was working on optimizing Goodbot, a program that plays Tantrix,
and because of the nature of the game, the only way to really qualify
an improvement is to run many test games against a standard opponent.
At one point, I was
I think the question is largely meaningless, because few games have
been studied by humans (or human computer programmers) with the depth
and intensity that has been achieved for games like chess and go.
In general, games with many choices and no obvious strategies
are good for people and bad
For those of you who use windows, I highly recommend tortoise cvs
and tortoise svn, which map access to whichever repository you prefer
into an incredibly useful and intuitive interface piggybacked on windows
explorer.
___
computer-go mailing list
For those of you who use windows, I highly recommend tortoise cvs
and tortoise svn, which map access to whichever repository you prefer
into an incredibly useful and intuitive interface piggybacked on windows
explorer.
___
computer-go mailing list
I suggest you add an identical RNG for testing purposes, which you
will know is identical in both implementations even if it is not
ideal. Run a test with a seeded random sequence which should
provide identical playouts.
___
computer-go mailing list
The formalized rules are the tortured details I referred to. I've
played thousands of games of Go, and I've never even seen any of those
versions of the rules.
The Japanese rules I refer to are the informal procedures I use every
time I play, both to estimate the score during the game, and at
Japanese: bad.
I don't think this is the case at all. The Japanese rules
are just a human optimization, to avoid having to make the
last 100 meaningless moves, and still arrive at the correct
score with a minimum of extraneous manipulation.
The tortured details, while not elegant, rarely
This was typically to pick up my queen, change its colour, and
capture my rook with it.
Now there's a feature that would make a tournament interesting...
If this appeals to you, try Martian Chess or Shogi
___
computer-go mailing list
I think the result
computer in hopelessly lost position resigns.
is much more satisfactory than
computer in hopelessly lost position wins by playing 100
additional pointless moves
I think a human who used this tactic in a tournament situation
might win the trophy, but would be unable to
I think the result
computer in hopelessly lost position resigns.
is much more satisfactory than
computer in hopelessly lost position wins by playing 100
additional pointless moves
I think a human who used this tactic in a tournament situation
might win the trophy, but would be unable to
I'm really very weak on networking so I'm not sure what I'm actually
reading or whether this fix needs to be applied on the server end or the
client end. Any ideas is this is relevant?
You also have the same problem, but with much less real information,
if the client end end of the connection
I'm really very weak on networking so I'm not sure what I'm actually
reading or whether this fix needs to be applied on the server end or the
client end. Any ideas is this is relevant?
You also have the same problem, but with much less real information,
if the client end end of the connection
I watched all the games, and I must say, mogo performed really badly
at the blitz games, and quite a bit better at the 1-hour game. I'd still
take any claims of dan level play with lots of salt.
My take-away from watching the match is that blitz performance wasn't
at all representative. A
1) Does anybody know of a good Java SGF parser out there?
I have one I've used for many types of games, including Go. I've used
it to represent large collections with no problems.
___
computer-go mailing list
computer-go@computer-go.org
1) Does anybody know of a good Java SGF parser out there?
I have one I've used for many types of games, including Go. I've used
it to represent large collections with no problems.
___
computer-go mailing list
computer-go@computer-go.org
Of course C can be more or less platform independent if you take some care.
Purely for engine code, that's true. Standard windows has APIs
that are nearly compatible with xxux for command line initialization
and ordinary file and network operations.
If your program has ANY gui at all
The thing that really kills multiple platform programs are GUIs.
Unless your GUI is born expecting to be cross platform, you're
pretty much screwed.
___
computer-go mailing list
computer-go@computer-go.org
By the way, has anyone seen the Philip Morris commercials?
I believe they were forced into this as part of the extortion
by the state attorneys general. It's Penance for illegally
targeting young non-smokers with Joe Camel, and promoting their
products while denying that they were
To a first order approximation, would changing the komi change the
rankings? Presumably, programs are playing the same number of games
as black and white, so any unfair advantage or disadvantage black
has would balance out.
Komi only matters when there is only one game between a pair of
At 11:44 AM 1/31/2008, David Doshay wrote:
That is correct.
It is my understanding that the Intel machines can compile to
a universal binary that will run on the G5 machines, but we
have not verified that. I trust that it works, but have no idea
if there is an efficiency hit.
Universal binaries
At 11:44 AM 1/31/2008, David Doshay wrote:
That is correct.
It is my understanding that the Intel machines can compile to
a universal binary that will run on the G5 machines, but we
have not verified that. I trust that it works, but have no idea
if there is an efficiency hit.
Universal binaries
It's interesting to look at a graphic plot of a traceroute
to see there the actual delays are. I use a program called
pingplotter for this, but there are many such programs.
Be warned though, that seeing a potential problem only leaves you
feeling helpless, since there is typically nothing you
CGOS uses Chinese scoring with play-outs so that we can get fully
automated scoring with no chance of errors.
No chance of errors is vacuously true. Errors, if any, were made
in the playout leading to the final state. There can be score
differences compared to what would have been Japanese
The standard one is Benson's algorithm
http://senseis.xmp.net/?BensonsAlgorithmhttp://senseis.xmp.net/?BensonsAlgorithm
The standard caveat is that this algorithm alone is very weak - it
typically applies to zero stones on a position played out using
Japanese rules. But you have to start
The standard one is Benson's algorithm
http://senseis.xmp.net/?BensonsAlgorithmhttp://senseis.xmp.net/?BensonsAlgorithm
The standard caveat is that this algorithm alone is very weak - it
typically applies to zero stones on a position played out using
Japanese rules. But you have to start
There's a sort of hierarchy of life-and-death methods, for which
Benson's algorithm is the base.
My status database is next above that, but it is actually a lookup table
based on a problem solver, such as Wolfs or mine. The unique thing
about the database is that it could be dropped in to a
My program was written in lisp, so naturally I concur. I'm not
actively using lisp any more, but I will offer various dialects
of common lisp as the consensus choice of dialect.
My favorite implementation is lispworks. The personal edition is
free and ought to be adequate for research.
The
At 05:24 AM 12/12/2007, Don Dailey wrote:
I've looked into this a bit. My preference would be scheme and it's
my understanding that it may be a bit more efficient.
If you're worried about efficient use of the machine, stay away from lisp
and scheme. Despite the claims of it can be as fast as
At 05:24 AM 12/12/2007, Don Dailey wrote:
I've looked into this a bit. My preference would be scheme and it's
my understanding that it may be a bit more efficient.
If you're worried about efficient use of the machine, stay away from lisp
and scheme. Despite the claims of it can be as fast as
These are true, but not the underlying problem.
The biggest underlying reason is the multiple constraints on
memory management;
a) since the data is typed rather than the pointers, every chunk of memory
has to be self identifying, not just for the garbage collector, but also so
(plus a
1 - 100 of 149 matches
Mail list logo
